supplies
Looking around in this thread you'll see a lot of experiments with electrolytic capacitors. But hardly anyone mentions the spikes originating from the rectifiers. There's a simple solution to that. Look at http://www.hawkaudio.nl/tips.htm
And even that can be improved by adding extra resistors and capacitors as in our D-402: http://www.hawkaudio.nl/D402.pdf
Any amp will perform better with a good power supply. A battery has the advantage that any disturbance originating from the wall outlet is eliminated. A disadvantge though is that you'll need very big batteries for a serious amp and those need to be loaded over and over again, a hassle.Daveis said:
Looking around in this thread you'll see a lot of experiments with electrolytic capacitors. But hardly anyone mentions the spikes originating from the rectifiers. There's a simple solution to that. Look at http://www.hawkaudio.nl/tips.htm
And even that can be improved by adding extra resistors and capacitors as in our D-402: http://www.hawkaudio.nl/D402.pdf
Re: supplies
Hello John,
Yes, the trick with the resistors works of course. Adding additional small inductors will further filter out HF rubbish (like the LC filter on the UcD modules).
Best regards
Gertjan
johnrtd said:
Hello John,
Yes, the trick with the resistors works of course. Adding additional small inductors will further filter out HF rubbish (like the LC filter on the UcD modules).
Best regards
Gertjan
Re: supplies
Hi John,
Those spikes are wrt to the currents which charge the resevoir caps to maintain peak voltage. I'm of the camp to optimize the charging path as much as possible, keep it fast, short path, minimal loop area, low noise (snubbed, more efficient than a series resistor as it only dissipates ringing energy from the switching instead of charging current), all for the sake of low ESR or ESZ.
..... that's one reason we like big toroids in an unregulated supply after all, and being such an expensive component, you want the most from it, even average rectifiers can handle the charging spikes. The real key to how good it gets is in whichever method you choose to seperate the input/charging currents with the output, and/or from channel to channel, which your implementation seems to concentrate on.
Regards,
Chris
johnrtd said:
Hi John,
Those spikes are wrt to the currents which charge the resevoir caps to maintain peak voltage. I'm of the camp to optimize the charging path as much as possible, keep it fast, short path, minimal loop area, low noise (snubbed, more efficient than a series resistor as it only dissipates ringing energy from the switching instead of charging current), all for the sake of low ESR or ESZ.
..... that's one reason we like big toroids in an unregulated supply after all, and being such an expensive component, you want the most from it, even average rectifiers can handle the charging spikes. The real key to how good it gets is in whichever method you choose to seperate the input/charging currents with the output, and/or from channel to channel, which your implementation seems to concentrate on.
Regards,
Chris
Re: Re: supplies
Small inductors are needed to supress "switching noise" from the UcD modules. If you want to improve the mains filtering you'll need rather big inductors, such as in tube amp supplies.
Those resistors are rather old fashioned. We found those sparks when developing a new amp and utilising a very fast oscilloscope. It came out that those rectifier spikes can cause detection in the first stage of an amp (or op amp). Upon detection you'll have an extra DC-voltage there and so on ....
Modern capacitors have a low impedance even at 100 kHz. So when adding a resistor the slope of the spike will get less steep and the capacitor does the rest. Two resistors in series and two capacitors improves this function of course.
After we applied the technique to all our designs we found that Philips is using the same trick in medical equipment and they were doing it for over 20 years!
Afterwards we tried the same trick in a lot of other amps and always with good results: less aggressive, smoother voices etc.
Dear Gertjanghemink said:
Small inductors are needed to supress "switching noise" from the UcD modules. If you want to improve the mains filtering you'll need rather big inductors, such as in tube amp supplies.
Those resistors are rather old fashioned. We found those sparks when developing a new amp and utilising a very fast oscilloscope. It came out that those rectifier spikes can cause detection in the first stage of an amp (or op amp). Upon detection you'll have an extra DC-voltage there and so on ....
Modern capacitors have a low impedance even at 100 kHz. So when adding a resistor the slope of the spike will get less steep and the capacitor does the rest. Two resistors in series and two capacitors improves this function of course.
After we applied the technique to all our designs we found that Philips is using the same trick in medical equipment and they were doing it for over 20 years!
Afterwards we tried the same trick in a lot of other amps and always with good results: less aggressive, smoother voices etc.
It came out that those rectifier spikes can cause detection in the first stage of an
HF is known to be rectified via op amp input stages, but if that HF on your amps' signal inputs has your PSU rectifiers to thank as a source, you've got far greater problems imho.
CRC or CRCRC is best reserved for a supply of light load for minimal droop under heavy load.. since there are no real heavy loads, like in the aux supply of an input stage. CLC ... where the L can just be a small inductor or common mode choke such as a ferrite clamp, to limite high frequency hash from say output stage HF ripple (whatever's left over after all the onboard filtering and bypassing), as well as any "remaining" ringing energy from your PSU rectifiers. It's best to qwell that ringing which you describe to be problematic in your tips area that quote from below, closer to the source, which is directly across the bridge, and maybe closer still through careful rectifier selection (softer/faster reverse recovery... low leakage. etc.) So the CLC will better allow the current to jam at high output with no frequency dependant droop imposed, no softening, relaxed "effect" by causing droop on the hard transients, which you should be supplying cleanly /stiffly instead for best, most natural, true to the source "effect".
For a fast, accurate, analytical, audiophile supply, one which can best take advantage of stereo imaging and soundstage by more accurately reproducing the input signal, I'd keep the charging path to the caps as low impedance across the highest band of frequencies as possible. Ensure a clean transmission by doing as stated above and you can further filter the mains accordingly.
You can also use ferrite clamps on inputs to reduce the actual RF HF noise that gets to the inputs.
Medical equipment has other concerns than an audiophile amplifier in some cases.
"This capacitor together with the inductance of the wiring causes a very sharp peak voltage to occur. This peak is too fast to be damped by the connected electrolytic capacitors."
" So when adding a resistor the slope of the spike will get less steep and the capacitor does the rest. "
What you're doing with the series resistor is creating a LP filter that will weight over the output. No doubt that has a smooth and relaxed effect but you're coloring the output by doing so and will suffer a frequency dependant droop under load as well.
I can tell you that by optimizing the charging path as described with the use of a more conventional high end supply (dual bridge rectifier with dual isolated secondaries, T networks (who recommends not to use a resistor here for audio, see Jensen white papers) produces the extremely clean, tightly imaged, , spacious and lifelike reproduction with a very black noise floor..
You pay a little more for parts that can do the job right though, but you're selling a kit of reduced cost.
HF is known to be rectified via op amp input stages, but if that HF on your amps' signal inputs has your PSU rectifiers to thank as a source, you've got far greater problems imho.
CRC or CRCRC is best reserved for a supply of light load for minimal droop under heavy load.. since there are no real heavy loads, like in the aux supply of an input stage. CLC ... where the L can just be a small inductor or common mode choke such as a ferrite clamp, to limite high frequency hash from say output stage HF ripple (whatever's left over after all the onboard filtering and bypassing), as well as any "remaining" ringing energy from your PSU rectifiers. It's best to qwell that ringing which you describe to be problematic in your tips area that quote from below, closer to the source, which is directly across the bridge, and maybe closer still through careful rectifier selection (softer/faster reverse recovery... low leakage. etc.) So the CLC will better allow the current to jam at high output with no frequency dependant droop imposed, no softening, relaxed "effect" by causing droop on the hard transients, which you should be supplying cleanly /stiffly instead for best, most natural, true to the source "effect".
For a fast, accurate, analytical, audiophile supply, one which can best take advantage of stereo imaging and soundstage by more accurately reproducing the input signal, I'd keep the charging path to the caps as low impedance across the highest band of frequencies as possible. Ensure a clean transmission by doing as stated above and you can further filter the mains accordingly.
You can also use ferrite clamps on inputs to reduce the actual RF HF noise that gets to the inputs.
Medical equipment has other concerns than an audiophile amplifier in some cases.
"This capacitor together with the inductance of the wiring causes a very sharp peak voltage to occur. This peak is too fast to be damped by the connected electrolytic capacitors."
" So when adding a resistor the slope of the spike will get less steep and the capacitor does the rest. "
What you're doing with the series resistor is creating a LP filter that will weight over the output. No doubt that has a smooth and relaxed effect but you're coloring the output by doing so and will suffer a frequency dependant droop under load as well.
I can tell you that by optimizing the charging path as described with the use of a more conventional high end supply (dual bridge rectifier with dual isolated secondaries, T networks (who recommends not to use a resistor here for audio, see Jensen white papers) produces the extremely clean, tightly imaged, , spacious and lifelike reproduction with a very black noise floor..
You pay a little more for parts that can do the job right though, but you're selling a kit of reduced cost.
Re: Re: Re: supplies
Hi John,
Thanks for the additional clarification. I was thinking more along lines similar as classd4sure, filtering out diode switching induced ringing. So pretty much HF noise. I have experimented with RCR type of filtering. I have to say that I did that with an SMPS, not with a conventional linear supply. In case of an SMPS, additional LC filtering with ferrite cores or common mode chokes made out of simple ferrite cores (DIY) works better for me as it removes more HF rubbish than an RC filter and indeed the impedance at lower frequencies is lower than with an RC setup. Of course the L values were quite small so that in the audio frequency range the impedance stays low. I agree with classd4sure, I also prefer to have a low impedance path between supply and amp at audio frequencies.
Thanks and best regards
Gertjan
johnrtd said:
Hi John,
Thanks for the additional clarification. I was thinking more along lines similar as classd4sure, filtering out diode switching induced ringing. So pretty much HF noise. I have experimented with RCR type of filtering. I have to say that I did that with an SMPS, not with a conventional linear supply. In case of an SMPS, additional LC filtering with ferrite cores or common mode chokes made out of simple ferrite cores (DIY) works better for me as it removes more HF rubbish than an RC filter and indeed the impedance at lower frequencies is lower than with an RC setup. Of course the L values were quite small so that in the audio frequency range the impedance stays low. I agree with classd4sure, I also prefer to have a low impedance path between supply and amp at audio frequencies.
Thanks and best regards
Gertjan
Hi guys 🙂
All this chat about PS is VERY interesting.
I have made "P filter" with my UCD400, as some may remember, and found sound improvement. I will try this R before bridges.
Yes, but, is the whole PS impedance that counts for "fast response" or only the path between the last cap ( i use 4*10000uF) and UCD module?
A scope can show freq. response of the amp, wright John? 😉
Thanks,
M
All this chat about PS is VERY interesting.
I have made "P filter" with my UCD400, as some may remember, and found sound improvement. I will try this R before bridges.
Yes, but, is the whole PS impedance that counts for "fast response" or only the path between the last cap ( i use 4*10000uF) and UCD module?
A scope can show freq. response of the amp, wright John? 😉
Thanks,
M
I'm not mr. Wright for sure! Indeed the last caps act as the main sources for big currents if an amp should reproduces a sudden staccato etc. Nice thing then is that one has a bit more power for a short time, say 10 ms, but the average power stays the same (when using R-C-R-C).maxlorenz said:
There's been quite a discussion here about different kinds of rectifiers, but IMHO all rectifiers show up the same rubbish. A good thing though is to choose good capacitors for the last stage in the power supply. I mean types as developed for use in switching systems. If money is a concern and if you're going to use a double filtering (as we do) you could choose a smaller value for the first capacitors. In most applications for the UcD180 and UcD400 a 4700 microfarad would be sufficient. I think, but never tried or measured, that even 1000 microfarad is enough.
I've been listening to and measuring amps with a far bigger capacitance (100.000 microfarad and up) and I regret to say they all had a "dull" sound.
A lot of people in this thread (and the UcD400) opt for bigger power. When listening to acoustical music the "life" experience is in small power, details of instruments and the acoustical signature of the recording hall are amplified at a milliwatt level. The nice thing with the UcD's is their low noise level. And when combined with a good power supply (avoiding disturbances from the wall outlet, dirty earth sources and rectifiers) you get a very "musical" amp.
And yes Max, one can measure the frequency response with an oscilloscope. But a good millivolt meter will do too or even better! And all this is NOT about frequency response or bandwidth.
Please post some diagrams of improvements you've made to the stock HG supply.
Also, has anyone replaced the AD opamp with a discrete version?
Also, has anyone replaced the AD opamp with a discrete version?
Hi Daveis
I think most of the discussions on power supplies here are from people who build their own. But ok, if I have missed something, I'm curious to find out. I use my HG-supply as is (so far).
If you want to replace the SMD-Opamp with a discrete circuit, you can remove the caps between the opamp and the modulator. This gives you a direct input path to the modulator, as decribed elsewhere in this forum. You could then have your discrete Opamp on a little pc-board, powered by the aux-supply of th HG-board, and preferably located close to XLR-inputs.
Kurt
I think most of the discussions on power supplies here are from people who build their own. But ok, if I have missed something, I'm curious to find out. I use my HG-supply as is (so far).
If you want to replace the SMD-Opamp with a discrete circuit, you can remove the caps between the opamp and the modulator. This gives you a direct input path to the modulator, as decribed elsewhere in this forum. You could then have your discrete Opamp on a little pc-board, powered by the aux-supply of th HG-board, and preferably located close to XLR-inputs.
Kurt
Thanks Johnrtd, for the reply🙂
Good to confirm my hypotheses...
In fact, now I'm not interested in big power but in clean power, because I'm moving to high efficiency speakers.
I bought Sikorel big cans for cheap. They are fast and detailed but maybe not as good as T-nets on high Freqs.
As I am unable to calculate Pi-filters, I was adviced by Rha61 to do:
10000uF//100uF---4uH---10000uH (with inductance having +/-0.5ohm).
I did not spent money on inductors and used a 10uH Epcos ferrite with N30 material (wich Epcos says is good for lower freqs) in series with a 0.5R. I heard more ellegant sound and better depth of image, without compromising transient response or HF extension. The 100uF cap is a cheap electro. When funds allow I will try, maybe, a BG FK in its place.
I have tried previously snubbing:
a) the bridges, 47nf+1ohm (maybe too low), and did not hear differences.
b) the last PS caps, and found no change (10nF+1R) or softer, rounder and slower sound (47nF+10R).
Regards,
M
Good to confirm my hypotheses...
In fact, now I'm not interested in big power but in clean power, because I'm moving to high efficiency speakers.
I bought Sikorel big cans for cheap. They are fast and detailed but maybe not as good as T-nets on high Freqs.
As I am unable to calculate Pi-filters, I was adviced by Rha61 to do:
10000uF//100uF---4uH---10000uH (with inductance having +/-0.5ohm).
I did not spent money on inductors and used a 10uH Epcos ferrite with N30 material (wich Epcos says is good for lower freqs) in series with a 0.5R. I heard more ellegant sound and better depth of image, without compromising transient response or HF extension. The 100uF cap is a cheap electro. When funds allow I will try, maybe, a BG FK in its place.
I have tried previously snubbing:
a) the bridges, 47nf+1ohm (maybe too low), and did not hear differences.
b) the last PS caps, and found no change (10nF+1R) or softer, rounder and slower sound (47nF+10R).
Regards,
M
"In fact, now I'm not interested in big power but in clean power, because I'm moving to high efficiency speakers."
There's absolutely no reason why the two can't co exist, but you have to look at each problem and go with the most direct cure that doesnt' simply address a symptom or two down the chain while blanketing eveything else with the same bandaide.
Be it the last cap in the chain or the first cap, they and every wire used on them is entirely audible.
Again the reason for the series resistor if used is to limite peak currents, _not_ ringing frequencies, and not anything do with mains frequencies either, these current charging spikes which the resistor limits only happen for a fraction of the mains period, and are high in peak value because of the high capacitance in the supply.
I've yet to see anyone myself attempt a 100 000uF supply for a UCD amp.... 10 000uF to 20 000uF is more normal and plenty high end.
"There's been quite a discussion here about different kinds of rectifiers, but IMHO all rectifiers show up the same rubbish"
Granted, but the better selected ones will show less rubbish /easier to snub, ask Hypex why they use schottky's.
"I've been listening to and measuring amps with a far bigger capacitance (100.000 microfarad and up) and I regret to say they all had a "dull" sound."
Try concentrating on unifying the charging currents and keeping them clean instead of limiting them and you'll at least retain some speed, if you manage to unify them via equal parasitics you'll obtain a tight/clear image. The level of capacitance must also be balanced with the level of VA available to charge it.
You get a good "musical" amp by employing a good /accurate ground scheme, whereby ground currents dont' disturb signal currents.
The "milliwatt" level microdetail you speak of typically rides on dozens of watts of more potent signal..... amps being musical intruments must accurately reproduce power from next to nothing to full on, as ideally as possible. This prevents the pampering of any particular aspect of the signal if you want a true reproduction.
" If money is a concern and if you're going to use a double filtering (as we do) you could choose a smaller value for the first capacitors. In most applications for the UcD180 and UcD400 a 4700 microfarad would be sufficient. I think, but never tried or measured, that even 1000 microfarad is enough."
The sound character of mixed caps add together, it's equally important to choose good ones throughout the chain so the final product is at least sonically pleasing at the end of the chain.
This includes the local bulk storage caps of the modules..... which are last in the chain, influence the sound heavily, and in stock modules, selected more for their cost than audible properties. (hint).
The reason for selecting a smaller capacitor as the first one in a CRC chain is not to save cost or anything else, it is again an attempt, much like the series resistor actually is, to get away with cheaper rectifiers by reducing the peak charging currents (not the ringing currents caused by them.... that's what snubbers are for). What you can get away here is in fact decided by the level of ripple current/life exptancy you're happy with.
Look at the modules and the Hypex supply though, aux input stage supply uses CRC filtering... light load. The modules employ CLC filtering on the rails.....
Gertjan, 😉
Regards,
Chris
There's absolutely no reason why the two can't co exist, but you have to look at each problem and go with the most direct cure that doesnt' simply address a symptom or two down the chain while blanketing eveything else with the same bandaide.
Be it the last cap in the chain or the first cap, they and every wire used on them is entirely audible.
Again the reason for the series resistor if used is to limite peak currents, _not_ ringing frequencies, and not anything do with mains frequencies either, these current charging spikes which the resistor limits only happen for a fraction of the mains period, and are high in peak value because of the high capacitance in the supply.
I've yet to see anyone myself attempt a 100 000uF supply for a UCD amp.... 10 000uF to 20 000uF is more normal and plenty high end.
"There's been quite a discussion here about different kinds of rectifiers, but IMHO all rectifiers show up the same rubbish"
Granted, but the better selected ones will show less rubbish /easier to snub, ask Hypex why they use schottky's.
"I've been listening to and measuring amps with a far bigger capacitance (100.000 microfarad and up) and I regret to say they all had a "dull" sound."
Try concentrating on unifying the charging currents and keeping them clean instead of limiting them and you'll at least retain some speed, if you manage to unify them via equal parasitics you'll obtain a tight/clear image. The level of capacitance must also be balanced with the level of VA available to charge it.
You get a good "musical" amp by employing a good /accurate ground scheme, whereby ground currents dont' disturb signal currents.
The "milliwatt" level microdetail you speak of typically rides on dozens of watts of more potent signal..... amps being musical intruments must accurately reproduce power from next to nothing to full on, as ideally as possible. This prevents the pampering of any particular aspect of the signal if you want a true reproduction.
" If money is a concern and if you're going to use a double filtering (as we do) you could choose a smaller value for the first capacitors. In most applications for the UcD180 and UcD400 a 4700 microfarad would be sufficient. I think, but never tried or measured, that even 1000 microfarad is enough."
The sound character of mixed caps add together, it's equally important to choose good ones throughout the chain so the final product is at least sonically pleasing at the end of the chain.
This includes the local bulk storage caps of the modules..... which are last in the chain, influence the sound heavily, and in stock modules, selected more for their cost than audible properties. (hint).
The reason for selecting a smaller capacitor as the first one in a CRC chain is not to save cost or anything else, it is again an attempt, much like the series resistor actually is, to get away with cheaper rectifiers by reducing the peak charging currents (not the ringing currents caused by them.... that's what snubbers are for). What you can get away here is in fact decided by the level of ripple current/life exptancy you're happy with.
Look at the modules and the Hypex supply though, aux input stage supply uses CRC filtering... light load. The modules employ CLC filtering on the rails.....
Gertjan, 😉
Regards,
Chris
spikes
Dear Max
If I understand what you did you now have a series coil and a resistor. When using wirewound resistors you already have a coil there! So try leaving out the extra coil.
The trouble with those spikes (and rubbish from the mains resp. earth connections) is that they may introduce detection. And when this happens the biasing of the transistor or tube in the amplifier changes hence the "character" of the amp changes.
All I'm saying is that a simple trick with just (2 x) 0,1 Ohm ahead from the bridge rectifier may give a smoother overall sound (in any amp). Any hobbyist can apply it for a few dollars.
When getting more serious it's a good thing of course to choose better rectifiers and capacitors. But all that has a price.
Wiring the (power to the) UcD's also has its effect. Good thing is to keep it as short as possible. A good thing too is using massive copper wire, best is the enamelled kind as used in transformers and loudspeaker coils.
Dear Max
If I understand what you did you now have a series coil and a resistor. When using wirewound resistors you already have a coil there! So try leaving out the extra coil.
The trouble with those spikes (and rubbish from the mains resp. earth connections) is that they may introduce detection. And when this happens the biasing of the transistor or tube in the amplifier changes hence the "character" of the amp changes.
All I'm saying is that a simple trick with just (2 x) 0,1 Ohm ahead from the bridge rectifier may give a smoother overall sound (in any amp). Any hobbyist can apply it for a few dollars.
When getting more serious it's a good thing of course to choose better rectifiers and capacitors. But all that has a price.
Wiring the (power to the) UcD's also has its effect. Good thing is to keep it as short as possible. A good thing too is using massive copper wire, best is the enamelled kind as used in transformers and loudspeaker coils.
UcD application
don't swear me if this has been asked already, topic too big to read now and I'm lazy 🙂
I was wondering if there are other applications which could use the UcD? or
is it strictly audio amplifier ?
don't swear me if this has been asked already, topic too big to read now and I'm lazy 🙂
I was wondering if there are other applications which could use the UcD? or
is it strictly audio amplifier ?
Re: UcD application
What answer do you expect, can you be a bit more specific of what you had in mind?
OK, I`ll give you one application, you can attach dummy loads to them (4 OHm high wattage resistors) pump a 180W sine wave in those resistsors (using the UcD of course to amplify that sinewave), you can then use the heat of the resistors as a heater, the amps themselves can not be used as a heater as you will find that they stay very cool, even when pumping that much power. 🙂
There are however cheaper ways to create a heater.
Best regards
Gertjan
rmsaudio said:
What answer do you expect, can you be a bit more specific of what you had in mind?
OK, I`ll give you one application, you can attach dummy loads to them (4 OHm high wattage resistors) pump a 180W sine wave in those resistsors (using the UcD of course to amplify that sinewave), you can then use the heat of the resistors as a heater, the amps themselves can not be used as a heater as you will find that they stay very cool, even when pumping that much power. 🙂
There are however cheaper ways to create a heater.
Best regards
Gertjan
non-audio
What about using a UcD (or four ones) as a DC-amp driving a DC-motor? For instance for in a car. The battery can be loaded using solar panels on the roof.
Also a blower on a pipe organ could be controlled etc. Then there's still "music in the air"!
What about using a UcD (or four ones) as a DC-amp driving a DC-motor? For instance for in a car. The battery can be loaded using solar panels on the roof.
Also a blower on a pipe organ could be controlled etc. Then there's still "music in the air"!
Re: UcD application
You could use it as part of a low noise psu for a high quality audio amp, say a UcD180rmsaudio said:
Re: Re: UcD application
This is actually one application that could be really good (have been thinking of it before). You could use two UcD400 modules, of course coupling caps need to be removed. Feed both UcD400 modules with the same powersupply (to prevent power supply pumping) and supply a stable 1.5-2V voltage (what about a 1.5V battery) at the input of both amps, on one amp on the + input, on the other amp on the - input, this will give you a +-30-40V rail voltage for each rail for one or more UcD180 amps. I do not know if the UcD400 is stable under such operating conditions, but guess it would be OK. This would be an extremely decadent power supply. There is however at least one drawback (besides it being expensive), The power supply (the UcD400 output) would have a 400kHz signal with relatively large amplitude superposed on the DC output. This is undesired I would say and can be reduced with an additional LC filter.
Maybe, very maybe, I`m going to try this one day for fun as I will have some spare UcD180 modules after all my UcD400 amps are put to use (I do not want to run any risk with my UcD400 amps).
Cheers
Gertjan
Hans L said:
You could use it as part of a low noise psu for a high quality audio amp, say a UcD180
This is actually one application that could be really good (have been thinking of it before). You could use two UcD400 modules, of course coupling caps need to be removed. Feed both UcD400 modules with the same powersupply (to prevent power supply pumping) and supply a stable 1.5-2V voltage (what about a 1.5V battery) at the input of both amps, on one amp on the + input, on the other amp on the - input, this will give you a +-30-40V rail voltage for each rail for one or more UcD180 amps. I do not know if the UcD400 is stable under such operating conditions, but guess it would be OK. This would be an extremely decadent power supply. There is however at least one drawback (besides it being expensive), The power supply (the UcD400 output) would have a 400kHz signal with relatively large amplitude superposed on the DC output. This is undesired I would say and can be reduced with an additional LC filter.
Maybe, very maybe, I`m going to try this one day for fun as I will have some spare UcD180 modules after all my UcD400 amps are put to use (I do not want to run any risk with my UcD400 amps).
Cheers
Gertjan